[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f8j3lgpXrRtGyh6u5zZVb8dkt1HVE7RZdEMezDNuvLY0":3},{"code":4,"msg":5,"data":6},200,"操作成功",{"id":7,"title":8,"content":9,"digest":10,"source":10,"coverPath":11,"thumbsCoverPath":12,"isTop":13,"isShow":14,"baseClick":13,"clickCount":15,"createTime":16,"typeId":17,"isNewest":18,"newsInfoTypeRespVo":19,"voiceUrl":22,"voiceSize":23,"taskId":24,"releaseTime":25,"titleEn":26,"contentEn":27,"voiceUrlEn":28,"taskIdEn":29,"voiceSizeEn":30},1365,"蚂蚁开源发布2025大模型全景图：中美AI开发路线分化，AI编程工具迎来爆发增长","\u003Cp>\u003Cstrong class=\"ql-lineHeight-1-75\" style=\"font-size: 18px; color: rgb(255, 153, 0);\">在2025Inclusion·外滩大会AI开源见解论坛上，蚂蚁开源与Inclusion AI携手推出的《全球大模型开源开发生态全景与趋势报告》2.0版本引发行业关注。这份基于开源社区百余日动态的报告，系统梳理了人工智能开源领域的技术演进与生态格局，为全球开发者提供了全景式参考。\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"undefined\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F09\u002Fc1d608c1391b4aefbd9d5a67601b715b\u002FAA1MwNq6.png\" width=\"undefined\" height=\"undefined\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">报告核心的大模型开源开发生态全景图2.0版本，首次将114个高关注度开源项目按技术领域分类，覆盖AI Agent与AI Infra两大方向共22个细分领域。数据显示，62%的开源项目诞生于2022年10月后，平均“年龄”仅30个月，印证了AI开源生态的快速迭代特征。这种技术活力在开发者分布上同样显著：约36万全球开发者中，美国贡献24%，中国占比18%，印度、德国、英国分别以8%、6%、5%位列其后，中美开发者合计贡献超四成核心代码。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">在开源策略上，中美厂商呈现明显分化。中国厂商更倾向开放权重模型，通过共享技术底座促进生态共建；美国头部企业则多采用闭源模式，构建技术壁垒。蚂蚁开源负责人王旭形象比喻：“这些开源项目如同数字积木，中国企业的共享策略让全球开发者能更自由地组合创新，这种开放性正为全球AI生态注入新动能。”\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">报告特别指出，AI编程工具已成为开源社区最活跃的领域。这类工具通过自动生成、修改代码，显著提升开发效率。从工具形态看，可分为“命令行工具”与“集成开发环境插件”两类：前者以Google的Gemini CLI为代表，凭借轻量化优势快速普及；后者如Cline，通过深度整合开发流程提供一站式体验。数据显示，2025年新推出的编程工具平均获得3万开发者Star关注，其中Gemini CLI开源3个月即突破6万星标，创下增长纪录。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">工具热潮背后，是开发者对“AI助手”的强烈需求。王旭团队分析发现，模型厂商多从命令行工具切入，注重技术底层优化；用户体验团\u003C\u002Fspan>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">队则优先开发IDE插件，强化开发流程整合。这种技术路线分化，正推动编程效率发生革命性变化。报告预测，随着大模型能力提升，程序员将更多聚焦创意设计与复杂问题解决，重复性编码工作可能逐步由AI工具承担，软件开发行业的分工模式或将因此重塑。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"undefined\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F09\u002F5266989109d24651b91e532afecc7b94\u002FAA1MwScc.png\" width=\"undefined\" height=\"undefined\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">论坛同期发布的2025大模型发展时间线全景图，以可视化方式呈现了全球主流厂商的模型发布轨迹。该图涵盖开放参数模型与闭源模型，标注了参数规模、模态类型等关键信息，清晰展现了中美企业在技术路线上的竞争态势。例如，中国厂商在多模态模型开发上进展迅速，美国企业则在模型推理能力优化方面持续突破。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">报告总结了当前大模型发展的五大趋势：中美开源闭源路线分化加剧、MoE架构推动参数规模化、强化学习提升模型推理能力、多模态模型成为主流方向、模型评价形成主观投票与客观评测双轨模式。这些趋势不仅反映了技术演进方向，也为开发者提供了战略决策参考。例如，MoE架构的普及使得模型参数规模突破万亿级成为常态，而多模态融合则要求开发者具备跨模态数据处理能力。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(187, 187, 187);\">【新闻来源】ITBEAR科技资讯 \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1MwEEG?ocid=msedgntp&pc=CNNDDB&cvid=68c7756ec461443fbe41a5296e8314c2&ei=30\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(187, 187, 187);\">https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1MwEEG?ocid=msedgntp&pc=CNNDDB&cvid=68c7756ec461443fbe41a5296e8314c2&ei=30\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(187, 187, 187);\">（本网转发此文章，旨在为读者提供更多的信息资讯，所涉内容不构成投资、消费建议。文章事实如有疑问，请与有关方核实，文章观点非本网观点，仅供读者参考。）\u003C\u002Fspan>\u003C\u002Fp>","","https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F09\u002F6febfe361e7943fabbb8a37795892dbb\u002FAI领域.jpg","https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F09\u002Fthumbs\u002F6febfe361e7943fabbb8a37795892dbb\u002FAI领域.jpg",0,1,38,"2025-09-16 10:05",2,false,{"id":17,"name":20,"enName":21},"芯位视野","Xinwei Vision","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3A1830e5f9-a8c0-4cae-8710-a9153a5c11b2%3A0.wav?Expires=1770036875&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=oSdSf4Pe4zhbE3Rwz28wEyxOAdI%3D",6693792,"1830e5f9-a8c0-4cae-8710-a9153a5c11b2","2025-09-16 10:01","Ant Financial Open Source Releases the 2025 Large Model Overview: Divergence in AI Development Paths Between China and the US, AI Programming Tools Experience Explosive Growth","\u003Cp>\u003Cstrong class=\"ql-lineHeight-1-75\" style=\"font-size: 18px; color: rgb(255, 153, 0);\">At the 2025 Inclusion·Outer Bank Conference AI Open Source Insights Forum, Ant Financial Open Source and Inclusion AI jointly released the \"Global Large Model Open Source Ecosystem Overview and Trend Report\" 2.0 version, which has attracted industry attention. Based on the dynamics of more than a hundred days of open source communities, this report systematically sorts out the technological evolution and ecological structure in the field of artificial intelligence open source, providing global developers with a comprehensive reference.\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"undefined\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F09\u002Fc1d608c1391b4aefbd9d5a67601b715b\u002FAA1MwNq6.png\" width=\"undefined\" height=\"undefined\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">The core large model open source development ecosystem overview 2.0 version of the report first classified 114 high-interest open source projects by technical fields, covering two major directions, AI Agent and AI Infra, with 22 sub-fields. Data shows that 62% of open source projects were born after October 2022, with an average age of only 30 months, verifying the rapid iteration characteristics of the AI open source ecosystem. This technological vitality is also significant in developer distribution: among about 360,000 global developers, the United States contributed 24%, China accounted for 18%, India, Germany, and the UK followed with 8%, 6%, and 5% respectively, and the combined contribution of Chinese and American developers exceeded 40% of the core code.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">In terms of open source strategy, Chinese and American companies show clear differentiation. Chinese companies tend to open weight models, promoting ecological co-construction through sharing the technical foundation; while leading US enterprises mostly adopt closed-source models, building technical barriers. Wang Xu, head of Ant Financial Open Source, made an image analogy: \"These open source projects are like digital building blocks, and the sharing strategy of Chinese companies allows global developers to freely combine innovations. This openness is injecting new momentum into the global AI ecosystem.\"\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">The report specifically points out that AI programming tools have become the most active area in the open source community. These tools significantly improve development efficiency by automatically generating and modifying code. From the tool form perspective, they can be divided into two categories: \"command line tools\" and \"integrated development environment plugins\": the former, represented by Google's Gemini CLI, quickly gained popularity due to its lightweight advantages; the latter, such as Cline, provides a one-stop experience by deeply integrating the development process. Data shows that in 2025, newly launched programming tools received an average of 30,000 developer stars. Among them, Gemini CLI surpassed 60,000 stars within three months of being open-sourced, setting a record for growth.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">Behind the tool craze is the strong demand from developers for \"AI assistants.\" Wang Xu's team found that model vendors often start with command line tools, focusing on optimizing the technical bottom layer; user experience teams prioritize developing IDE plugins, enhancing the integration of the development process. This technical route divergence is driving a revolutionary change in programming efficiency. The report predicts that as large model capabilities improve, programmers will focus more on creative design and complex problem solving, and repetitive coding tasks may gradually be taken over by AI tools, potentially reshaping the division of labor in the software development industry.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cimg alt=\"undefined\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F09\u002F5266989109d24651b91e532afecc7b94\u002FAA1MwScc.png\" width=\"undefined\" height=\"undefined\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">At the same time, the 2025 large model development timeline overview was released during the forum, presenting the model release trajectories of global mainstream manufacturers in a visualized way. This chart covers open parameter models and closed-source models, marking key information such as parameter scale and modal type, clearly showing the competitive situation between Chinese and American companies in technology routes. For example, Chinese manufacturers have made rapid progress in multi-modal model development, while American companies continue to make breakthroughs in model reasoning capability optimization.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\" class=\"ql-lineHeight-1-75\">The report summarizes five trends in current large model development: the intensifying divergence of open source and closed source routes between China and the US, MoE architecture driving parameter scaling, reinforcement learning enhancing model reasoning ability, multi-modal models becoming the mainstream direction, and the formation of a dual-track model evaluation system combining subjective voting and objective evaluation. These trends not only reflect the direction of technological evolution but also provide strategic decision-making references for developers. For instance, the popularization of MoE architecture has made it common for model parameters to exceed ten billion, while multi-modal integration requires developers to possess cross-modal data processing capabilities.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(187, 187, 187);\">【News Source】ITBEAR Technology News \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1MwEEG?ocid=msedgntp&pc=CNNDDB&cvid=68c7756ec461443fbe41a5296e8314c2&ei=30\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(187, 187, 187);\">https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Far-AA1MwEEG?ocid=msedgntp&pc=CNNDDB&cvid=68c7756ec461443fbe41a5296e8314c2&ei=30\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(187, 187, 187);\">（This article is reprinted by this site to provide readers with more information and news. The content involved does not constitute investment or consumption advice. If there are any questions about the facts of the article, please verify with the relevant parties. The views of the article are not the views of this site and are for reference only.)\u003C\u002Fspan>\u003C\u002Fp>","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3Aed659795-a4d1-46bd-b373-0c22cef37a49%3A0.wav?Expires=1774838470&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=aBpLX%2BXTBtrJeliTBSu6T88ujdo%3D","ed659795-a4d1-46bd-b373-0c22cef37a49",10087242]